ANTICANCER RESEARCH 36: 3065-3070 (2016)

Distance from Is the Strongest Predictor for Early Onset of Brain Metastases in

ESZTER GORKA1, DÁNIEL FABÓ2, ANDRÁS GÉZSI3, KATA CZIRBESZ1 and GABRIELLA LISZKAY1

1Department of Dermato-, National Institute of Oncology, Budapest, Hungary; 2Epilepsy Centre, Department of Neurology, National Institute of Clinical Neurosciences, Budapest, Hungary; 3Department of Genetics, Cell and Immunbiology, Semmelweis University, Budapest, Hungary

Abstract. Background/Aim: The frequency of brain Therefore, the aim of this retrospective study was to identify (BM) is up to 45-50% in patients with advanced melanoma. risk factors useful in predicting the early development of BM Our aim was to identify the risk factors for the early and to highlight the strongest ones in order to determine a occurrence of BM. Patients and Methods: A total of 333 high-risk patient group. patients with BM were identified from our database of 2,972 patients with melanoma between 2003-2015. Results: The Patients and Methods median elapsed time to BM (TTBM) was significantly associated with Breslow thickness, ulceration, location, and From our melanoma database of 2,972 patients between 2003 and patient age. Head and neck location was the strongest predictor 2015, we identified 333 diagnosed with BM. One group of these patients was treated at our Center from the discovery of their primary for early BM development [hazard ratio (HR)=1.81, 95% melanoma, the other was referred to us after the development of BM confidence interval (CI)=1.05-3.12; p=0.031) followed by for complex oncotherapy, including stereotactic radiosurgery. The Breslow thickness >2 mm (HR=1.53, 95% CI=1.04-2.23; following parameters were registered: gender, age at diagnosis p=0.027). Body part-specific median TTBM was 51.5, 43, 38.5, (tested in categories per 10 years), location of the primary melanoma 32, 35, 36.5, 35.5 and 19 months in leg-foot, thigh, abdomen- [head and neck (HN), trunk, upper extremity and lower extremity], pelvic, chest-back, lower arm-hand, upper arm-shoulder, face- Breslow thickness (investigated in two categories of <2.0 mm and neck and scalp regions, respectively. Conclusion: We suggest ≥2.0 mm), ulceration (presence or absence), histological subtype (nodular, superficial spreading, acral lentiginous or lentigo maligna brain magnetic resonance imaging follow-up in the high-risk melanoma) and sentinel (SLN) status (positive for patient group of patients with melanoma in the head and neck metastasis or negative). SLN was performed mostly in cases region, especially for those with primary melanoma over of intermediate tumor thickness excluding the HN region, according Breslow 2 mm located in the scalp. to our institutional protocol from 1999. We analyzed these parameters in association with the median Brain metastases (BM) are frequent in stage IV malignant elapsed time to brain metastasis (TTBM). TTBM was calculated melanoma: brain involvement has increased to 45-50% in from the date of primary tumor diagnosis to the detection date of BM. With those factors showing significant correlation with TTBM, those having advanced melanoma (1) and 75% at autopsy further analysis was carried out to compare their predictive value. (2). Even with the use of new targeted and immune therapies, Finally, we investigated the distance-based associations by BM is still the most important factor as it determines quality distributing the primary site regions into smaller categories and then of life and life expectancy. However, in cases of early determining TTBM for each. detection, the results are promising. Regarding the new therapeutic possibilities, early detection has an increasing Statistical methods. After having described the whole study role in managing patients with melanoma with BM. population, we excluded patients with melanoma whose primary was unknown. For TTBM analysis, patients whose primary melanoma was not removed before they first presented with BM were also excluded. In order to compare the primary site to brain distances objectively along a straight axis, there was further Correspondence to: Eszter Gorka, Department of Dermato- exclusion of patients with upper limb melanoma. For mapping all oncology, National Institute of Oncology, Rath Gyorgy u. 7-9, 1122, distance-related associations, we analyzed every region again. Budapest, Hungary. Tel: +36 12248600 2519, Fax: +36 12248727, The effect of the variables on early BM occurrence was e-mail: [email protected] determined by univariate Cox regression analysis, and the combined effect of these predictors was assessed by using multivariate Cox Key Words: melanoma, brain metastases, predictive factors, location, regression analysis. The significance of the models was evaluated distance, scalp, head and neck. by the log-rank test and results with two-sided p<0.05 were

0250-7005/2016 $2.00+.40 3065 ANTICANCER RESEARCH 36: 3065-3070 (2016) considered significant. All analysis was performed in R statistical Table I. Characteristics of the melanoma study population. software (R Foundation for Statistical Computing, Vienna, Austria; version 3.0.3.) using the survival package in R. Kaplan–Meier Total cases, n (%) 333 (100) curves, were plotted using the survMisc package in R. Gender, n (%) Male 196 (59) Female 137 (41) Results Mean age (range), years 53.9 (13.5-86) Site of primary melanoma, n (%) A total of 333 patients with melanoma were included in our Total known 283 (85) Head and neck region 46 (16.3) analysis, who developed BM during the study period (2003- Trunk 145 (51.2) 2015). Patient characteristics are shown in Table I. The Upper extremity 38 (13.4) median Breslow thickness was 3.0 mm (mean=4.13 mm, Lower extremity 54 (19.1) range=0.1-40 mm), 173 patients (69.5%) fell in the ≥2 mm Unknown primary 50 (15) and 76 patients (30.5%) into the <2 mm groups. Mean Breslow thickness (range), mm 4.13 (0.1-40) Ulceration, n (%) The median TTBM for the whole patient group was 36 Total known 195 (59) months (range=0.5-305 months). Potential risk factors were Yes 113 (58) analyzed in association with TTBM, starting with the No 82 (42) location of the primary melanoma. The median TTBM was No data 138 (41) 32 months for primary of the HN region, 34.5 months for Histological subtype, n (%) Total known 171 (52) those of the trunk, while it was much longer, 51.5 months, Nodular 97 (56.7) for those of the lower limb region (Figure 1). Compared with Superficial spreading 67 (39.2) melanoma located in the lower limb, location in the trunk Acral lentiginous 4 (2.3) conferred a significantly higher risk of shorter TTBM Lentigo maligna 3 (1.8) [hazard ratio (HR)=1.4, 95% confidence interval (CI)=1.01- No data 161 (48) SLNB, n (%) 83 (25) 1.94; p=0.046;N=190], and the HN location conferred even Positive 23 (28) greater increased risk for early onset of BM (HR=2.18, 95% Negative 60 (72) CI=1.44-3.32; p=0.00026; N=95). Detailed analysis of the Median TTBM (range), months 36 (0.5-305) primary location (Figure 2) showed that the median TTBM was 51.5 months when melanoma started in the leg-foot SLNB: Sentinel ; TTBM: time to brain metastasis. region (n=36, 13%), 43 months if disseminated from the thigh (n=12, 4%), 38.5 months from the abdomen-pelvic region (n=38, 14%), 32 months from the chest-back region (n=79, 29%), 35 months from the lower arm-hand region ulceration (as previously detailed). We found no significant (n=8, 3%), 36.5 months from the upper arm-shoulder (n=54, association between TTBM and gender (p=0.17), 20%), 35.5 months from the face and neck (n=24, 9%), and histological subtype (p=0.53) or SLN status (p=0.30). was only 19 months for those originating in the scalp region We carried out further analysis with the significant (n=22, 8%). In this respect, scalp location was sharply independent risk factors using a multivariate Cox regression separated from the other parts of the HN (HR=1.84, 95% model including age, location, tumor thickness and ulceration CI=1.01-3.34; p=0.047) and showed outstanding risk for BM to evaluate their predictive strength for TTBM. As a reference compared to those arising from the leg-foot region point of the comparison, we took a hypothetical low-risk group (HR=3.03, 95% CI=1.76-5.22; p=0.000063). The face-neck of young patients with thin (<2 mm) primary melanomas region was also a significant risk factor compared with the located on the lower extremity, without ulceration, and then the leg-foot regime (HR=1.75, CI=1.03-2.96; p=0.038). joint effect of the studied predictors (six variables: age, location In the group with Breslow thickness ≥2 mm, the median of HN, of upper limb and of trunk, Breslow of ≥2 mm, and TTBM was 31 months, while for those with <2 mm it was presence of ulceration) was compared to this group. Our results 47.5 months (HR=1.37, 95% CI=1.05-1.8; p=0.023; n=249). showed that HN location (HR=1.81, 95% CI=1.05-3.12; In cases of ulceration, the median TTBM was 26.5 months, p=0.031) and Breslow thickness ≥2 mm (HR=1.53, 95% while in the non-ulcerated group it was 43 months (HR=1.58, CI=1.04-2.23; p=0.027) remained statistically significant out CI=1.17-2.13; p=0.0025; n=195). As for age, we found that of six variables, and the HN location was the strongest older age elevated the risk of short TTBM (per 10-year predictor among all, followed by Breslow tumor thickness increase: HR=1.19, CI=1.09-1.31; p=0.00014; n=272). (Figure 3). When Breslow thickness was neglected in this TTBM was significantly correlated with patient age, the calculation, ulceration became significant, and the second primary location (either as a single variant, by log-rank test strongest factor after HN location. The association between p=0.001), the Breslow thickness and the presence of tumor thickness and ulceration may explain this finding.

3066 Gorka et al: Role of Primary Site in Melanoma Progression to Brain

Figure 1. Kaplan–Meier curves of distance-based time to brain metastasis (TTBM). The risk of early brain metastasis significantly drops in the following order of primary tumor location: head and neck, trunk, and lower extremity.

Figure 2. Detailed subdivision of primary tumor location and time to brain metastasis (TTBM). TTBM tends to gradually decrease with proximity to the scalp. Note that scalp location is sharply demarcated from other head and neck regions.

Discussion any dissemination is diagnosed, BM can already be detected in 20% of all cases (3). Early detection can prolong survival BM development is a frequent complication of malignant because of better surgical and radiosurgical options for fewer melanoma. The prognosis depends on the staging categories and smaller lesions. The median overall survival time after of M1a, M1b, M1c, even in disseminated phase. By the time developing BM is generally 4 months, but radiosurgery can

3067 ANTICANCER RESEARCH 36: 3065-3070 (2016)

Figure 3. Relative risk associated with the main variables for early onset of brain metastasis from melanoma. Head and neck location is the strongest predictor for early onset, followed by greater Breslow tumor thickness. HR: Hazard ratio; CI: confidence interval.

result in 6.6 months, which might be even improved to 12.5 Bottoni et al.'s mean tumor thickness of 4.0 mm (8) and months by combining it with surgical resection in selected approached the results of Timothy et al., as they found a mean population (4). For patients with stable BM, more favorable Breslow thickness of 3.4 mm and an ulceration rate of 55% results are expected with the possibility of new therapies. for their BM population (6), but surprisingly differed from the Targeted and immunotherapies statistically improved survival 1.85 mm tumor thickness and 48% ulceration of Zakrzewski et in clinical studies, even in cases of cerebral progression. al. (9). Since our study population selectively included patients Median overall survival reached 33 weeks with dabrafenib (4) with brain metastases, we analyzed the variables in accordance and 7 months with ipilimumab in cases of asymptomatic BM with the TTBM, while other studies mostly calculated BM (5), while nivolumab and pembrolizumab are currently being frequency; thereby we identified the predictors of early BM explored in several ongoing trials in patients with BM from onset instead of predictors of BM incidence. Our collected melanoma (5). Early detection of BM and the use of new TTBM data of 36 months is longer than those of the literature therapies would maximize the chance for the best survival. data [23-30.5 months (7-9)], but if by not excluding cases with There have been several prior studies aimed at finding BM unknown primaries, we obtained 28 months as the TTBM that predictors in patients with melanoma. In these studies, is similar to the results of these studies. Breslow thickness and ulceration were reported to be the TTBM was significantly correlated with older age, HN major risk factors, while certain studies added HN location, location of the primary tumor, greater Breslow depth and the age, gender, SLN status and histological melanoma subtype presence of ulceration, consistent with prior studies. to prior predictors (6-8). The predictive strength of these However, we found no significant association between variables is controversial, but some authors suggest that TTBM and gender, histological subtype or SLN status, location and ulceration are the strongest ones, consequently contrary to certain reports where male gender (10-11), acral patients with primary ulcerated melanoma on the head and lentiginous and nodular primary tumor (12), or positive SLN neck might be at higher risk for development of BM (9). (8) were concluded as risk factors for BM. Even with our Moreover, within the HN region, primary scalp 28% sentinel positivity rate, which is higher than literature were recently reported to be associated with a higher data of 23.8% (8), and our previously reported data of 14.7% incidence of BM than face or neck melanoma (10). (13), there was no significance regarding TTBM. Our study includes a large dataset of patients with We observed that the distance between the primary melanoma with BM compared to prior studies, where the melanoma site and the brain strongly affected TTBM. number of the BM population mostly remained under 150 Accordingly, we analyzed our data in a distance-based patients. The demographic and clinicopathological features of manner instead of the usual axial (HN, trunk) versus the current analysis were mostly similar to previously peripheral (limbs) site distribution, although axial location of published data. Our patients' mean Breslow thickness of primary melanoma is widely considered to be more likely to 4.13 mm and the ulceration rate of 58% corresponded to develop BM than peripheral. Our results demonstrated that

3068 Gorka et al: Role of Primary Site in Melanoma Progression to Brain the risk of early BM was higher in HN than trunk-located 3 Long GV, Trefzer U, Davies MA, Kefford RF, Ascierto PA, melanoma, which was at higher risk than location in the Chapman PB, Puzanov I, Hauschild A, Robert C, Algazi A, lower extremity. The large number of patients in the current Mortier L, Tawbi H, Wilhelm T, Zimmer L, Switzky J, Swann S, Martin AM, Guckert M, Goodman V, Streit M, Kirkwood JM analysis enabled us to perform more specific subdivision of and Schadendorf D: Dabrafenib in patients with Val600Glu or regions regarding the anatomical site of the primary tumor. Val600Lys BRAF-mutant melanoma metastatic to the brain In this way, the scalp location resulted in a TTBM of 19 (BREAK-MB): a multicentre, open-label, phase 2 trial. Lancet months, hence sharply demarcated from other HN regions. Oncol 11: 1087-1095, 2012. This face and neck region falls under the standard category 4 Christ SM, Mahadevan A, Floyd SR, Lam FC, Chen CC, Wong ET of the trunk and upper extremity level, with TTBM in the and Kasper EM: Stereotactic radiosurgery for brain metastases range of 32-38.5 months. As a next lower segment, the thighs from malignant melanoma. Surg Neurol Int 20: 6, 2015. were associated with a median TTBM of 43 months, while 5 Ramanujam S, Schadendorf D and Long GV: Systemic therapies for melanoma brain metastases: which drug for whom and the lowest segment including legs and feet had the longest, when? Chin Clin Oncol 4(2): 25, 2015. 51.5 months median TTBM. Consequently, a distance-based 6 Frankel TL, Bamboat ZM, Ariyan C, Coit D, Sabel MS and subdivision appears to be preferable instead of axial versus Brady MS: Predicting the development of brain metastases in peripheral distribution. patients with local/regional melanoma. J Surg Oncol 109: 770- 774, 2014. Conclusion 7 Quian M, Ma MW, Fleming NH, Lackaye DJ, Hernando E, Osman I and Shao Y: Clinicopathological characteristics at primary melanoma diagnosis as risk factor for brain metastasis. Our study revealed that the HN region, and especially Melanoma Res 23(6): 461-467, 2013. location in the scalp, was the strongest predictor for BM 8 Bottoni U, Clerico R, Paolino G, Ambrifi M, Corsetti P and onset among all factors, the second prognostic factor was a Calvieri S: Predictors and survival in patients with melanoma Breslow thickness >2 mm, followed by the presence of brain metastases. Med Oncol 30: 466, 2013. ulceration and older age respectively. Based on these 9 Zakrzevszky J, Geraghty LN, Rose AE, Christos PJ, Mazumdar M, findings, we identified a high-risk group of patients for BM Polsky D, Shapiro R, Berman R, Darvishian F, Hernando E, Pavlick with HN (scalp) located primary melanoma greater than A and Osman I: Clinical variables and primary tumor characteristics Breslow thickness of 2 mm. For this well-circumscribed predictive of the development of melanoma brain metastases and post-brain metastases survival. 117: 1711-1720, 2011. relatively small subpopulation, we suggest a closer 10 Huismans AM, Haydu LE, Shannon KF, Quinn MJ, Saw RP, evaluation using brain MRI which has not generally been Spillane AJ, Stretch JR and Thompson JF: Primary melanoma part of routine follow-up protocols in clinical practice. In this location on the scalp is an important risk factor for brain way, BM would be discovered at an earlier stage and could metastasis: a study of 1687 patients with cutaneous head and be treated more effectively, particularly when in possession neck melanomas. Ann Surg Oncol 21: 3985-3991, 2014. of new therapies. 11 Daryanani D, Plukker JT, de Jong MA, Haaxma-Reiche H, Nap R, Kuiper H and Hoekstra HJ: Increased incidence of brain Acknowledgements metastases in cutaneous head and neck melanoma. Melanoma Res 15: 119-124, 2005. 12 Sampson JH, Carter JH, Friedman AH and Seigler HF: This study was supported by the Hungarian government Demographics, prognosis, and therapy in 702 patients with brain (KTIA_NAP 13-1-2013-0001). Parts of these results were presented metastases from malignant melanoma. J Neurosurg 88: 11-20, as a poster at the European Society for Medical Oncology 1998. Conference 2015 in Vienna. 13 Liszkay G, Orosz Z, Péley G, Csuka O, Plótár V, Sinkovics I, Bánfalvi T, Fejõs Z, Gilde K and Kásler M: Relationship References between status and regression of primary malignant melanoma. Melanoma Res 15(6): 509-513, 2005. 1 Barnholtz-Sloan JS, Sloan AE, Davis FG,Vigneau FD, Lai P and Sawaya RE: Incidence proportions of brain metastases in patients diagnosed (1973 to 2001) in the Metropolitan Detroit Cancer Surveillance System. J Clin Oncol 22(14): 2865-2872, 2004. 2 Sloan AE, Nock CJ and Einstein DB: Diagnosis and treatment Received April 5, 2016 of melanoma brain metastasis: a literature review. Cancer Revised April 29, 2016 Control 16(3): 248-255, 2009. Accepted May 11, 2016

3069